Thread Safety: LazyLock<reqwest::Client> + tokio::test

[sebadob]

I like to use reqwest::Client inside a LazyLock<reqwest::Client> to be able to access it easily from lots of different locations in applications and make proper use of the internal connection pooling and keep alives, instead of recreating the client each time again which would be a big waste of ressources on TLS handshakes.

This client is not behind any Mutex and I am using it directly - everything seems to be fine and I did not encounter any issues with this in real running applications. However, when I tried to add some #[tokio::test] fn's that use this client, it seems that the answers it parses are getting mixed up, when tests run in parallel. As soon as I run them sequential, the issues go away.

I know that the tokio test runtime does some things to internal timers and so on, so maybe it's just my test runtime setup, but I wanted to make sure, that LazyLock<reqwest::Client> is safe to use without locking in a "real" runtime - is this the case, or should I be worried, as soon as traffic and parallel requests increase?

[seanmonstar]

reqwest's Client is safe to use across threads.

You might find problems if you try to use it in multiple runtimes, because of it holds onto connections and timers, and those only work if the runtime that created them is still running.

[sebadob]

You might find problems if you try to use it in multiple runtimes, because of it holds onto connections and timers, and those only work if the runtime that created them is still running.

Yeah then this is probably what my tkoio::test setup does under the hood.

Thank you very much!